Unadhi Guides

"Why was I refused?" Credit score and AI score explanation sheet for accountants, tax advisers and their SME clients (GDPR Articles 15(1)(h), 22, C-634/21 SCHUFA, C-203/22 Dun & Bradstreet Austria)

Updated 2 Nov 2026 · Disclosure: the EURO AI Register mentioned below is an Unadhi product with a free edition. This guide is for information only, not legal advice.

What you get: a finished credit score and AI score explanation sheet for accountants, tax advisers, bookkeepers and the small business clients they advise: online shops with pay-later checks, landlords and letting agents, wholesalers, lenders, insurers' brokers and employers using AI CV screeners. One row per score or automated decision: type in twelve short answers and the sheet returns whether it is likely an Article 22 decision, the Article 22(2)-(4) status (OK, GAP or STOP), the reply-by date for an access request (one month, Article 12(3)), a countdown (DUE IN n DAYS, URGENT, DUE TODAY, LATE BY n DAYS) and a to-do list citing the GDPR article or Court of Justice judgment behind each step. Also included: a 10-question checklist, a plain-language explanation template, a reply letter to the person, an email to the credit agency or AI supplier and a client email. Formulas tested in LibreOffice on ten made-up rows.

Short answer: since C-634/21 SCHUFA (7 December 2023), a credit agency's automatically calculated probability score is itself an "automated individual decision" under GDPR Article 22(1) where a third party "draws strongly" on it to establish, implement or terminate a contract with the person. Since C-203/22 Dun & Bradstreet Austria (27 February 2025), the person can require the controller to explain "the procedure and principles actually applied" to their data, "in a concise, transparent, intelligible and easily accessible form" (Article 15(1)(h)). Sending a complex mathematical formula or an algorithm, or describing every step, is not enough (para 59). Trade secrets are not a reason to refuse outright: the controller has to give the withheld information to the data protection authority or court, which balances the interests (ruling 2).

Why this lands on the accountant's desk

Step 1: tab 1, Scores (one row per score or automated decision)

ColumnWhat to writeSource
AScore or automated decision (for example "pay-later check at checkout")You
BWho calculates the score: Us or Supplier (credit agency, AI tool)C-634/21
CUsed for: Credit or payment terms, Hiring, Pricing, Fraud check, Marketing or otherRecital 71
DDoes a human review each case and can they change the outcome? Y or NArt 22(1) "solely"
EHow often the outcome follows the score: Always, Almost always or SometimesC-634/21 "draws strongly", para 48
FEffect on the person: Refusal or end of contract, Worse price or terms, Job rejection, Account blocked or Ad or offer onlyArt 22(1), recital 71
GBasis for the decision: Contract, Law, Explicit consent, Consent or NoneArt 22(2)
HHealth or other Article 9(1) data used? Y or NArt 22(4)
IIs the person a natural person (consumer, applicant, sole trader)? Y or NArt 1, 4(1)
JAccess request received on (date; blank if none)Art 12(3)
KPlain explanation of the factors ready? Y or NArt 15(1)(h), C-203/22
LHuman review and contest route offered? Y or NArt 22(3)
MArticle 22 decision? (our reading): =IF(A2="","",IF(I2="N","NO: NOT A NATURAL PERSON",IF(NOT(OR(F2="Refusal or end of contract",F2="Worse price or terms",F2="Job rejection",F2="Account blocked")),"PROFILING: LIKELY NOT ART 22 (OUR READING)",IF(OR(D2="N",E2="Always",E2="Almost always"),IF(D2="Y","LIKELY ART 22: HUMAN REVIEW DOES NOT CHANGE THE OUTCOME (OUR READING)","ART 22 DECISION"),"NOT SOLELY AUTOMATED: KEEP THE HUMAN REVIEW REAL"))))Art 22(1), C-634/21
NArticle 22(2)-(4) status: =IF(M2="","",IF(NOT(OR(LEFT(M2,3)="ART",LEFT(M2,6)="LIKELY")),"N/A",IF(AND(H2="Y",G2<>"Explicit consent",G2<>"Law"),"STOP: SPECIAL CATEGORY DATA (ART 22(4))",IF(G2="None","STOP: NO ART 22(2) EXCEPTION",IF(G2="Law","CHECK THE SAFEGUARDS IN THAT LAW (ART 22(2)(b))",IF(L2<>"Y","GAP: OFFER HUMAN INTERVENTION, VIEW, CONTEST (ART 22(3))","OK: EXCEPTION AND SAFEGUARDS"))))))Art 22(2)-(4)
OReply by: =IF(OR(A2="",J2=""),"",EDATE(J2,1))Art 12(3)
PRequest status (Settings!B1 holds today: =TODAY()): =IF(A2="","",IF(J2="","NO REQUEST",IF(O2-Settings!$B$1<0,"LATE BY "&(Settings!$B$1-O2)&" DAYS",IF(O2=Settings!$B$1,"DUE TODAY",IF(O2-Settings!$B$1<=3,"URGENT: "&(O2-Settings!$B$1)&" DAY(S) LEFT","DUE IN "&(O2-Settings!$B$1)&" DAYS")))))Art 12(3)
QTo-do: =IF(M2="","",TEXTJOIN("; ",TRUE,IF(LEFT(M2,3)="NO:","the GDPR protects people, not companies: check sole traders, partners and named directors before you close this row",""),IF(AND(OR(LEFT(M2,3)="ART",LEFT(M2,6)="LIKELY"),K2<>"Y"),"write a plain explanation of the procedure and principles actually applied: which of the person's data were used, how they counted and what change would have led to a different result; a formula or algorithm alone is not enough (Art 15(1)(h), C-203/22 paras 58-62)",""),IF(AND(OR(LEFT(M2,3)="ART",LEFT(M2,6)="LIKELY"),B2="Supplier"),"a supplier score you draw strongly on is itself an Art 22 decision of the supplier (C-634/21): ask the supplier for its Art 15(1)(h) explanation and the person's profile",""),IF(N2="STOP: NO ART 22(2) EXCEPTION","stop the solely automated decision or add a real human review: Art 22(2) allows it only if necessary for a contract, authorised by law or with explicit consent",""),IF(LEFT(N2,14)="STOP: SPECIAL","no solely automated decision on health, ethnic origin or other Art 9(1) data unless explicit consent or substantial public interest law and safeguards (Art 22(4))",""),IF(LEFT(N2,3)="GAP","give the person at least a way to get human intervention, express their view and contest the decision (Art 22(3))",""),IF(OR(LEFT(M2,3)="ART",LEFT(M2,6)="LIKELY"),"say in the privacy notice that the automated decision exists, the logic, the significance and the consequences (Art 13(2)(f), 14(2)(g))",""),IF(LEFT(M2,3)="NOT","log who reviewed each case, what they looked at and when they overrode the score",""),IF(LEFT(M2,9)="PROFILING","access requests still get the data, purposes and source (Art 15(1)); the logic duty in Art 15(1)(h) applies at least to Art 22 cases",""),IF(AND(J2<>"",LEFT(M2,3)<>"NO:"),"send a copy of the data used, including the score itself (Art 15(1), 15(3)) by the reply date: one month, extendable by two further months with reasons (Art 12(3))",""),IF(AND(J2<>"",OR(LEFT(M2,3)="ART",LEFT(M2,6)="LIKELY")),"trade secrets or other people's data are no reason to refuse outright: the withheld information goes to the data protection authority or court, which balances the interests (C-203/22, ruling 2)",""),"write down who decided, why and the date"))see each item

In Google Sheets or Excel, paste each formula into row 2 and fill down. Format J and O as dates. "One month" is computed with EDATE (same day next month); that is our reading of how to count it, so check national practice for weekends and holidays.

Tested rows (made up, Settings!B1 fixed at 2 Nov 2026)

RowKey answersMNP
1: online shop pay-later check, agency score auto-refuses, request 20 OctSupplier, no human, contract, no review routeART 22 DECISIONGAP: OFFER HUMAN INTERVENTION, VIEW, CONTEST (ART 22(3))DUE IN 18 DAYS
2: landlord tenant screening, agent always follows the score, request 31 OctHuman "reviews" but almost always follows, no basisLIKELY ART 22: HUMAN REVIEW DOES NOT CHANGE THE OUTCOME (OUR READING)STOP: NO ART 22(2) EXCEPTIONDUE IN 28 DAYS
3: wholesaler credit limit for limited companiesNot a natural personNO: NOT A NATURAL PERSONN/ANO REQUEST
4: AI CV screener rejects applicants below 60, request 3 OctUs, no human, no basisART 22 DECISIONSTOP: NO ART 22(2) EXCEPTIONURGENT: 1 DAY(S) LEFT
5: AI fraud tool, staff review each blockHuman reviews, outcome sometimes changesNOT SOLELY AUTOMATED: KEEP THE HUMAN REVIEW REALN/ANO REQUEST
6: newsletter segment by AI purchase score, request 1 NovAd or offer onlyPROFILING: LIKELY NOT ART 22 (OUR READING)N/ADUE IN 29 DAYS
7: gym AI price from a health questionnaireHealth data, contract basisART 22 DECISIONSTOP: SPECIAL CATEGORY DATA (ART 22(4))NO REQUEST
8: sole-trader loan pre-check, explicit consent, review route, request 5 OctExplanation readyART 22 DECISIONOK: EXCEPTION AND SAFEGUARDSURGENT: 3 DAY(S) LEFT
9: risk flag required by a tax law, request 2 OctLaw basisART 22 DECISIONCHECK THE SAFEGUARDS IN THAT LAW (ART 22(2)(b))DUE TODAY
10: broker AI quote refusal, request 25 Sep ignoredSupplier, almost always, contract, review routeART 22 DECISIONOK: EXCEPTION AND SAFEGUARDSLATE BY 8 DAYS

The to-do for row 10, for example, starts: "write a plain explanation of the procedure and principles actually applied: which of the person's data were used, how they counted and what change would have led to a different result; a formula or algorithm alone is not enough (Art 15(1)(h), C-203/22 paras 58-62); a supplier score you draw strongly on is itself an Art 22 decision of the supplier (C-634/21)..." An empty row returns nothing.

Step 2: the 10-question checklist (one score at a time)

  1. Is there a decision? The Court reads "decision" broadly; it can include the result of calculating a person's creditworthiness as a probability value (C-634/21 para 46). Recital 71 names "automatic refusal of an online credit application or e-recruiting practices without any human intervention".
  2. Is it based solely on automated processing? Ask how often staff actually change the outcome. If the answer is "almost never", record why the review is still real or treat it as automated (our reading).
  3. Does it have legal or similarly significant effects? Refusal of a contract, worse terms, a job rejection, a blocked account. A marketing segment usually does not (our reading).
  4. Is the score from a supplier? If you draw strongly on it, the supplier's score is itself an Article 22 decision (C-634/21, ruling). Ask the supplier how it handles Article 15(1)(h) requests.
  5. Which Article 22(2) exception applies? Necessary for a contract, authorised by Union or Member State law with safeguards, or explicit consent. If none, stop the solely automated decision.
  6. Is health or other Article 9(1) data used? Then only explicit consent or substantial public interest law, with safeguards (Article 22(4)).
  7. Can the person get human intervention, express their view and contest? Required at least for the contract and consent cases (Article 22(3)).
  8. Does the privacy notice say so? Articles 13(2)(f) and 14(2)(g) require the existence of the automated decision, meaningful information about the logic, and the significance and envisaged consequences.
  9. Is a plain explanation ready? Which data, how they counted, and, per C-203/22 para 62, possibly the extent to which a change in the data would have led to a different result.
  10. Is a request open? Reply within one month of receipt; extension by two further months only where necessary, with reasons given within the first month (Article 12(3)). Include a copy of the data, including the score (Article 15(1) and (3)).

Plain explanation of the score (template)

How we reached the decision on [date]: We used [tool / agency] to [assess whether to offer pay-later / set your credit limit / shortlist applicants]. It used these data about you: [list, for example payment history with us, postcode, number of open invoices, years at address]. The factors that counted most were: [factor 1, direction], [factor 2, direction], [factor 3, direction]. The result was [score / category], below our threshold of [x]. If [factor] had been [different value], the result would likely have been [different outcome]. [A staff member did / did not] review the result. You can ask for a person to review it, tell us your point of view and contest the decision by replying to this letter.

Reply letter to the person (template)

Subject: Your request of [date]: your data and how our decision was made

Dear [name],

Thank you for your request received on [date]. Attached is a copy of the personal data we used, including the score [value] we received from [agency / tool] on [date], and where we obtained them. The explanation of how the decision was made is below. [Paste the plain explanation.] If you think any data are wrong, tell us and we will check and correct them. You can ask for a person to review the decision and you can contest it. You also have the right to complain to [data protection authority].

Kind regards, [name, business]

Email to the credit agency or AI supplier (template)

Subject: [customer / applicant ref]: explanation of your score under GDPR Article 15(1)(h)

Hello,

We use your score in our decisions about [credit / tenancy / applicants]. Following Court of Justice judgments C-634/21 and C-203/22, please confirm by [date]: (1) how a person can obtain from you a concise explanation of the procedure and principles actually applied to their data; (2) the main factors your score used for the person in our reference [ref], in plain words we can pass on; (3) whether you will withhold any part as a trade secret, and if so, that you will provide it to the data protection authority or court on request; (4) the data you hold on the person and their sources.

Kind regards, [name, business]

Email to business clients (template)

Subject: Credit checks, pay-later and AI screening: one sheet to check them

Dear [name],

If you refuse customers, set payment terms, screen tenants or shortlist applicants using a score from an agency or an AI tool, the EU Court of Justice has said that such a score can count as an automated decision, and people can ask you how it was made. We have prepared a one-row-per-tool sheet that shows where you need a review route, a privacy notice line or a reply by a deadline. Please send us the list of scores and tools you use and any "why was I refused?" emails you have received.

This is a practical note, not legal advice. Kind regards, [firm]

Related free guides

Keep each score, its supplier and its evidence in one place

Record each scoring or AI tool, its supplier, the Article 22 answer and the evidence (supplier answers, privacy notice line, review route) in the EURO AI Register free edition. It runs in your browser with no account and no upload, covers your first 10 AI systems and 10 evidence records, the screening questions, duties by role with dates, and a printable Article 4 AI-literacy record.

When a client has more than 10 systems, the EURO AI Register membership starts at $19 a month and adds unlimited systems, rule updates and email help. Neither tool is legal advice or proof of compliance.

Partner programme opening soon

Partner programme opening soon: join the free waitlist. Planned terms are 50% of each referred sale (75% once you have 5+ paying customers), single level, paid through Gumroad only from revenue received.

Limits and open points

Read for this guide (2 Nov 2026, EUR-Lex): GDPR Articles 12(3), 13(2)(f), 14(2)(g), 15(1)-(4), 22(1)-(4) and recital 71; the rulings of C-634/21 and C-203/22 and paras 43-48 and 61-63 of C-634/21 and paras 58-64 and 66 of C-203/22.

Not read for this guide: the EDPB-endorsed guidelines on automated decision-making and profiling (WP251) themselves (cited only as quoted in C-203/22 para 60), the AI Act (including Annex III point 5(b) on creditworthiness and Article 86), the Consumer Credit Directive (EU) 2023/2225, national credit-agency and data protection laws under Article 22(2)(b) or 23, national authority guidance and Directive (EU) 2016/943 beyond its mention in the ruling.

Our reading, not the Court's words: column M's "likely" and "rubber stamp" grades, the thresholds "Always / Almost always" for "draws strongly", treating a marketing segment as not significant, the EDATE one-month count and the sample explanation wording. The Court left the final assessment in both cases to the national court.

Frequently asked questions

Is a credit agency score an automated decision even if we make the final call?

The Court of Justice held that the automated establishment of a probability value by a credit information agency is an automated individual decision under Article 22(1) where the third party it is sent to draws strongly on it to establish, implement or terminate a contract (C-634/21, 7 December 2023). Whether you draw strongly on it depends on how often your outcome simply follows the score.

Do we have to give out the scoring algorithm?

No. The Court said the mere communication of a complex mathematical formula, such as an algorithm, or a detailed description of all the steps is not a sufficiently concise and intelligible explanation (C-203/22, para 59). The person must be able to understand which of their data were used and in what way (para 61).

Can we refuse because the scoring method is a trade secret?

Not on your own. Where the controller considers the information contains trade secrets or other people's data, it must provide that information to the competent supervisory authority or court, which balances the rights and interests to decide how much the person gets (C-203/22, ruling 2).

How long do we have to answer?

Without undue delay and in any event within one month of receipt of the request. The period may be extended by two further months where necessary, taking into account complexity and number of requests, and the person must be told of the extension and the reasons within the first month (GDPR Article 12(3)).

Not legal advice. The sheet helps sort scores tool by tool; it does not decide legal questions. Spotted an error? Write to mininglife7@gmail.com and we correct it.

Sources

← All guides

Unadhi Guides are free, independent explainers. We correct errors as soon as we find them: write to mininglife7@gmail.com.